Prognostic and predictive tools are needed to optimize treatment for women diagnosed with ductal carcinoma in situ (DCIS). While radiotherapy (RT) is the standard of care for DCIS after breast-conserving surgery (BCS), those at low risk (≤10%) for ipsilateral breast recurrence (IBR) risk may be treated without RT. Historically, “low risk” has been defined as the absence of “high risk” clinicopathological factors, which include younger age (<50 yrs), tumors ≥2 cm, palpable tumors, or high nuclear grade (G) tumors. All prospective trials demonstrate that RT reduces IBR by ~50%, even in low-risk patients (RTOG 9804). In this study, we analyzed a cohort of women with DCIS treated with BCS +/– RT to determine if the biosignature could identify a subset of women who do not benefit from RT and evaluate the biosignature in patients meeting “low risk” or “high risk” clinicopathological criteria.
DCIS patients (n = 926) from 4 international cohorts (median follow-up, 8.5 yrs; 1st-3rd quartile, 5.8-10.2 yrs) treated with BCS (negative margins) with RT (n = 641) and without RT (n = 335), were evaluated for favorable clinicopathological criteria (age>50 or grade 1-2, and RTOG 9804–like disease [G1-2, screen detected, negative margins]) and clinical outcomes. Formalin-fixed, paraffin-embedded tissue samples for each patient were analyzed at a CLIA lab for the 7-gene biosignature and a residual risk subtype (RRt). The biosignature reported a decision score (DS) of 0-10 and the presence/absence of the RRt. A low-risk group (DS≤2.8 without RRt) was compared with the combined elevated risk (DS>2.8 without RRt) and residual risk groups (DS>2.8 with RRt), where 10-yr total IBR rates were evaluated using Cox proportional hazards and Kaplan-Meier analysis by treatment, biosignature risk group, and clinicopathological criteria.
The biosignature classified 37% of women treated with BCS as low-risk (n = 338) and 63% (n = 588) into the combined elevated/residual risk group. Among patients who did not receive RT, those in the elevated/residual risk group had higher IBR rates (P <.001) than those in the low-risk group with corresponding 10-yr IBR rates of 25.7% (95% CI, 18.8%- 34.4%) vs 5.6% (95% CI, 2.5%-12.1%), respectively. RT did not reduce IBR in the low-risk group (HR, 0.8; P = .71), where the 10-yr IBR rate was 4.8% (95% CI, 2.5%-9.1%) with RT and 5.6% without RT, corresponding to a number needed to treat (NNT) of ~100. The elevated/residual risk group benefited from RT (HR, 0.2; P < .001), with an absolute reduction of 14.5% (10-yr IBR, 8%), for a NNT of 6. About 50% of patients defined as clinicopathological -low-risk by 9804-like (51%) or favorable Age/G (58%) criteria were reclassified by the biosignature to elevated/residual risk. These pts had significant RT benefit: 9804-like group: HR 0.3; P = .007; absolute 10-yr IBR reduction of 12.7%; favorable Age/G group: HR, 0.34; P = .012, absolute 10-yr IBR reduction of 11.2%. The corresponding NNTs were ~8-9. For patients in clinicopathological high-risk groups, 23% of not-9804-like and 31% of age<50 or G3 pts were reclassified as biosignature low-risk. RT did not significantly reduce IBR in any low-risk Biosignature group, including those in clinicopathological high-risk groups. IBR for not-9804-like group was 5.9% vs 3.0% without and with RT (P = .52), and for (age<50 or G3) pts was 4.4% vs 3% without and with RT (P = .70).
The biosignature identified a low-risk group with excellent outcomes and no significant RT benefit even in those with G3 disease or aged <50 yrs. 10-yr IBR rates with or without RT remained independent of clinicopathological criteria, further supporting the identification of a true low-risk group who may forgo RT. In contrast, biosignature elevated/RRt patients had clinically meaningful IBR benefits with RT, even for those in favorable clinicopathologic groups (10-yr IBR reduction of 13%-26% with RT). These patients should not be offered RT omission, and these data highlight the lack of accuracy of clinicopathological factors in assessing RT benefit. The 7-gene biosignature provided a superior prediction of 10-yr risk and RT benefit than standard clinicopathological risk models.
Rachel Rabinovitch,1 Frank A. Vicini,2 Chirag Shah,3 Julie A. Margenthaler,4 Brian Czerniecki,5 Pat Whitworth,6 David Dabbs,7 G Bruce Mann,8 Fredrik Wärnberg,9 Sheila Weinmann,10 Michael Leo,10 Jess Savala,7 Steven C. Shivers,7 Karuna Mittal,7 Troy Bremer7
1University of Colorado Cancer Center, Aurora, CO.
2GenesisCare, Farmington Hills, MI.
3Cleveland Clinic Taussig Cancer Institute, Cleveland, OH.
4Washington University School of Medicine, St Louis, MO.
5Moffitt Cancer Center, Tampa, FL.
6Nashville Breast Center, Nashville, TN.
7PreludeDx, Laguna Hills, CA.
8University of Melbourne, Royal Women’s Hospital, Parkville, Australia.
9University of Gothenburg, Gothenburg, Sweden.
10Kaiser Permanente Center for Health Research, Portland, OR.